Information Extraction and Similarity Computation for Semi-/Un-Structured Sentences from the Cyberdata

Abstract

With the popularization of network and the improvement of network speed, social network application plays an increasingly important role in people’s social life. People express their opinions and ask their own questions on social software, and these huge amounts of data drive researchers to propose various algorithms to extract the information in sentences and classify them. In this paper, we proposed a novel method of sentence similarity computation, which purpose is to extract the syntactic and semantic information of semi-structured and structured sentences and calculate their similarity. We mainly consider the subject predicate and object of sentence pairs, and use Stanford parser to classify Dependency Relation Triples to calculate the syntactic and semantic similarity between two sentences. Extensive simulations demonstrated that our method outperforms the other state-of-the-art methods in terms of correlation coefficient and mean deviation.